✓ CRACKED — STATE-READABLE · NO VISION NEEDED
FT09
FT09 · CLICK ONLY · 6 LEVELS · WIN ✓
Sprite attributes human-named (gig, fhc, gqb). Win condition fully readable from Python state.
85 agent actions vs 208 human baseline. Step budget conserved. All 6 levels cleared.
This is the proof of concept. Read the ground truth, act optimally, beat the human.
115%
RHAE SCORE
◑ WORMHOLE_NEEDED — CLICK-ONLY · STEP BUDGET BLOCKING · VISION = UNLOCK
LP85
WATCHING
step budget exhausted · need goal position mapping from rendered frame
R11L
WATCHING
move direction unclear without visual · target positions need render
TN36
WATCHING
correct button sequence unknown without seeing board state
VC33
WATCHING
placement validity requires visual — ezbubuphlm() check needs seen state
S5I5
WATCHING
target-to-slider mapping requires visual spatial reasoning
SU15
WATCHING
lphwskxbh drag tracking — correct drag sequence needs visual
✗ DEEP_SOLVE — KEYBOARD + COMPLEX STATE · VISION + PLANNING REQUIRED
SB26
213 actions
CD82
171 actions
TR87
414 actions
TU93
462 actions
G50T
879 actions
LS20
776 actions
AR25
748 actions
BP35
651 actions
CN04
789 actions
DC22
1228 actions
KA59
730 actions
LF52
1339 actions
M0R0
1107 actions
RE86
1255 actions
SC25
350 actions
SK48
1070 actions
SP80
518 actions
WA30
1843 actions
⬡ THE VISION LAYER — RH1 IN THE HARNESS
The unlock for all 24 blocked games. One universal interface exists across every ARC-AGI-3 game:
Send to minicpm-v (running on msi01 Ollama, RTX integrated, no cloud needed):
"What needs to move where to solve this puzzle? Give me target x,y coordinates."
Execute the click. Read the new state. Repeat until WIN.
No H100. No cloud. No external API. Local inference, local execution, sovereign results.
This is RH1 in the harness. The rendered pixel frame IS the ground truth — the same truth the human sees. WATCHING → STRONG. The vision layer is the next sprint target.
frame_pixels = env._game.camera.render() → numpy array → the exact pixel frame the human sees.Send to minicpm-v (running on msi01 Ollama, RTX integrated, no cloud needed):
"What needs to move where to solve this puzzle? Give me target x,y coordinates."
Execute the click. Read the new state. Repeat until WIN.
No H100. No cloud. No external API. Local inference, local execution, sovereign results.
This is RH1 in the harness. The rendered pixel frame IS the ground truth — the same truth the human sees. WATCHING → STRONG. The vision layer is the next sprint target.
⬡ ALL 25 GAMES · FULL TOPOLOGY TABLE
| GAME | INPUT | LEVELS | HUMAN | AGENT | RHAE | STATUS | BLOCKER / NOTE |
|---|---|---|---|---|---|---|---|
| FT09 | click | 6 | 208 | 85 | 115% | WIN | State-readable · sprite attrs human-named · PROOF OF CONCEPT |
| LP85 | click | 8 | 388 | — | — | WATCHING | goal/goal-o tag visible · step budget · vision layer = unlock |
| R11L | click | 6 | 233 | — | — | WATCHING | sys_click sprites · sliding puzzle · target direction needs render |
| TN36 | click | 7 | 317 | — | — | WATCHING | Maidxz button grid · bltjrl pattern · correct sequence = visual |
| VC33 | click | 7 | 447 | — | — | WATCHING | pick/place mechanic · ezbubuphlm() validation = visual |
| S5I5 | click | 8 | 638 | — | — | WATCHING | slider resize puzzle · target-to-slider mapping = visual |
| SU15 | click+7 | 9 | 361 | — | — | WATCHING | drag tracking lphwskxbh · 224 action positions pre-loaded |
| SB26 | kb+click | 8 | 213 | — | — | DEEP | actions 5,6,7 · dewwplfix[8] board · animation state machine |
| CD82 | kb+click | 6 | 171 | — | — | DEEP | pixel-match: xytrjjbyib=eoqnvkspoa · yrfgxhebei() targets |
| TR87 | keyboard | 6 | 414 | — | — | DEEP | sequence matching · cifzvbcuwqe[6] pattern pairs |
| TU93 | keyboard | 9 | 462 | — | — | DEEP | 0017→0015 exit navigation · kdkehgjrzq 3-state machine |
| G50T | keyboard | 7 | 879 | — | — | DEEP | player+hazard physics · ACTION5=special · hazard moves every 2 |
| LS20 | keyboard | 7 | 776 | — | — | DEEP | position+rotation checkpoints · 3 rotate modes · nav puzzle |
| AR25 | kb+click | 8 | 748 | — | — | DEEP | dqwpuqcubca:21 · all 7 actions · complex multistate |
| BP35 | kb+click | 9 | 651 | — | — | DEEP | actions 3,4,6,7 · physics animation · 9 levels |
| CN04 | kb+click | 6 | 789 | — | — | DEEP | hlxyvcmpk+iahpylgry dicts · kpgnbcoir string state |
| DC22 | kb+click | 6 | 1228 | — | — | DEEP | deajuohco[2] · timing counter · ancwrndvh accumulator |
| KA59 | kb+click | 7 | 730 | — | — | DEEP | dynamic spawn · empty lists at start · hrknegnjkg grows |
| LF52 | click+kb | 10 | 1339 | — | — | DEEP | foztuzaztfm[10] · 10 levels · 1339 actions · complex |
| M0R0 | kb+click | 6 | 1107 | — | — | DEEP | anfcrclwoac dynamic · jnyxgyktgcv None→Sprite spawn |
| RE86 | keyboard | 8 | 1255 | — | — | DEEP | win_score:8 · 5 actions · xikvflgqgp string state machine |
| SC25 | kb+click | 6 | 350 | — | — | DEEP | bcffdwcmlef[32]+bmmtkvkbcdd[9] grid · cmagnysjqzl dynamic |
| SK48 | kb+click | 8 | 1070 | — | — | DEEP | lgdrixfno position tracker · jdojcthkf dict · full kb+click |
| SP80 | kb+click | 6 | 518 | — | — | DEEP | dkvpswzsjg:'change' state · avangppqui bool · set tracking |
| WA30 | keyboard | 9 | 1843 | — | — | END BOSS | ALL keyboard · 9 levels · 1843 actions · hardest game. Last. |